Speedtest CLI accurate?

[ccihon]

Recently when running Speedtest CLI from the webgui (or cron) I get what I think are accurate download speeds represented for my 1.2 fiber connection at always around 1250Mbps, but the upload speed, only recently, seems to top out as a measurement at less than 1000. The lan and wan interfaces are both at 2.5G, so it's not that anything should limit it to gigabit. Oddly, a downstream wired device connected through a switch etc to the lan can achieve 1200 up - so should I just chalk this up to a Speedtest CLI inaccuracy when run on the router? Thanks.

OPNsense 24.7.11_2-amd64
FreeBSD 14.1-RELEASE-p6

Speedtest by Ookla 1.2.0.84 (ea6b6773cf) FreeBSD 14.1-RELEASE-p6 amd64

[HardTack5]

It's not a fair test running the speed test on the router itself. The proper way would be to use a wired device connected to the router via the switch (or whatever).

It's one thing ROUTING traffic at speed. It's another thing sourcing and receiving the traffic at speed. It's not the same. Traffic to and from a router itself is not the same as traffic THROUGH a router. So yes, the test itself running on the router will change the result.

Speedtest CLI is pretty good.

[dseven]

I think running the test on the router/firewall itself is quite reasonable - think of it as being as close as possible to a "raw" test (without eliminating the firewall entirely), and it would eliminate variables like LAN/WiFi. You would, of course, want to ALSO do tests on clients behind the firewall, and compare the results. It seems reasonable to expect that the test on the firewall itself shouldn't show worse performance than on a routed client, but you would want to make sure that the tests are "apples to apples" - i.e. using the same test server(s), packet sizes, concurrent streams, etc....